铝合金穿孔型等离子弧立焊的BP神经网络预测模型

BP Neural Network Predicting Model for Aluminium Alloy Keyhole Plasma Arc Welding in Vertical Position

基于MATLAB6.1的神经网络工具箱 ,利用BP神经网络建立铝合金穿孔型等离子弧立焊的输入 -输出的网络模型 ,通过训练该网络 ,能够根据输入节点数值预测输出结果。当给定网络输入节点的各个焊接工艺参数值 ,能够预测网络的输出 ,即焊缝形状的各个参数。用这个工艺的匹配值进行焊接试验。结果表明 ,焊缝的形状参数与网络模型的预测结果之间的误差在 8%以内。仿真试验的结果表明 ,这个方案是可行的。

In this paper, based on MATLAB6.1 neural network toolbox, a BP neural network modal for vertical position PAW input-output is established.According to each value of input nodes,the output can be predicted by testing this network.When welding parameters of input nodes are given,the parameters of welding formation can be predicted.The experimental results made with the combination of each parameters show that the error between parameters of real welds and its predicting results is within 8 percent. The results of simulation experiments show that this way is practical.